Library¶

In [ ]:
import pandas as pd
import sys
sys.path.append('../../../Util')
import ShowGraphs as sg
from bertopic import BERTopic

1 Baseline Summary¶

Clustering Approach¶

  • Parameter Setting: A high parameter was set for HDBSCAN to ensure well-defined clusters that occupy a significant percentage of the total dataset. This baseline is intended to feed machine learning algorithms for prediction purposes.

Initial Clustering Results¶

  • Clusters Retrieved: 7 representative clusters were identified:

    1. Drug sales
    2. Bitcoin
    3. Scammers and seller reviews
    4. Marketplace advertising
    5. Purchase reviews
    6. Drug purchases
    7. Orders

  • Outliers: Initially, 34k outliers were found out of a total of 66k records.

  • Performance Metrics:

    • Silhouette Score: 0.64
    • Davies-Bouldin Score: 0.6
    • Coherence-Score: 0.40

Outlier Reduction¶

  • Cosine Measure on Embeddings: By applying a cosine similarity measure with a 0.53 threshold, the number of outliers was reduced from 34k to 27k, reintroducing about 7k records.

  • Updated Performance Metrics:

    • Silhouette Score: 0.51
    • Davies-Bouldin Score: 0.8
    • Coherence-Score: 0.40

Trade-off Analysis¶

  • Outlier Reintroduction: Reintroducing the outliers found a balance that prevented significant cluster degradation while keeping clusters well separated and defined, as evidenced by the graphs.

  • Cluster Distribution: The updated clusters are well-distributed:

    • Maximum cluster size: 23% of the total dataset
    • Minimum cluster size: 7% of the total dataset
    • This distribution avoids large excursions.

Data Loss and Potential Adjustments¶

  • Data Loss: Approximately 40% of the initial dataset was lost.

  • Potential Correction: This data loss can potentially be mitigated by lowering the cosine similarity threshold between embeddings.

Visualize Graphs CSV Files¶

In [2]:
topic_model = BERTopic.load("../../../Analyze_files/CombiningAnalysisCompleteDataset/ThreadAnalysis/Models/topic_model_0.64SilNew", embedding_model='distiluse-base-multilingual-cased-v1')
In [3]:
descr_topic = pd.read_csv('CSV7Topic/description_topic.csv')
document_topic = pd.read_csv('CSV7Topic/document_topic_proba.csv')
topics_over_time = pd.read_csv('CSV7Topic/topic_over_time_3.csv')

Topic Description¶

In [4]:
descr_topic
Out[4]:
Topic Count BERTopic_Name Representation Representative_Docs Custom_Name_GenAI Custom_Name_Zero_Shot
0 4 6299 4_mdma_lsd_get_looking ['mdma' 'lsd' 'get' 'looking' 'wsm' 'good' 'bt... ['hey ro im gon na pull pk'\n 'life wonderful ... Drug Discussion buy drugs
1 2 6907 2_vendor_scammer_scam_looking ['vendor' 'scammer' 'scam' 'looking' 'scamming... ['looking good reliable vendor sell ounce' 've... Trusted Vendor Scams scammer - vendor scammer
2 1 6236 1_market_empire_dream_nightmare ['market' 'empire' 'dream' 'nightmare' 'vendor... ['next market' 'dream market vendor rstclass n... Dream Vendor Nightmare market drug - empire market
3 5 2776 5_order_package_shipping_delivery ['order' 'package' 'shipping' 'delivery' 'pack... ['order accepted day still hasnt marked shippe... Order Shipping & Tracking orders
4 6 2823 6_bitcoin_card_wallet_btc ['bitcoin' 'card' 'wallet' 'btc' 'bank' 'credi... ['credit cards paypal prepaid card find'\n 'go... Financial Services and Products bitcoin wallet
5 0 8645 0_weed_xanax_vendor_cocaine ['weed' 'xanax' 'vendor' 'cocaine' 'mg' 'uk' '... ['high quality weed thc product europe'\n 'new... Drug Vendor Europe sell drugs
6 3 4230 3_review_vendor review_vendor_review vendor ['review' 'vendor review' 'vendor' 'review ven... ['empire vendor cocaine review' 'first ever re... Vendor MDMA Review review vendor - review product
In [5]:
sg.plot_topic_distribution(descr_topic)
No description has been provided for this image
In [6]:
sg.plot_topic_percentage_distribution(descr_topic)
No description has been provided for this image
In [7]:
sg.create_wordclouds(topic_model, num_topics=7, cols=2)
No description has been provided for this image

Document Topic Description¶

In [8]:
print(document_topic.shape[0])
document_topic.head(5)

37916
Out[8]:
Document Topic Probability Created_on BERTopic_Name
0 checks 4 0.000000 2020-01-09 4_mdma_lsd_get_looking
1 trusted vendor status 2 0.944247 2020-01-09 2_vendor_scammer_scam_looking
2 empire exit scam iiflux user incomming 1 1.000000 2019-11-06 1_market_empire_dream_nightmare
3 ecstasy vendor packs 5 0.797741 2020-01-09 5_order_package_shipping_delivery
4 opening bank account person fake id 6 1.000000 2019-11-06 6_bitcoin_card_wallet_btc
In [10]:
sg.plot_avg_prob_or_freq(document_topic)
No description has been provided for this image
In [11]:
sg.plot_boxplot(document_topic)
No description has been provided for this image
In [12]:
sg.plot_probability_distribution(document_topic)
No description has been provided for this image
In [13]:
sg.create_wordclouds(document_topic, num_topics=7, cols=2, is_model=False)
No description has been provided for this image
In [14]:
sg.create_topic_network_graphs(document_topic, max_words=80, min_occurrences=10)
No description has been provided for this image

Topics Over Time¶

In [15]:
print(topics_over_time.shape[0])
topics_over_time.head(5)

21
Out[15]:
Topic Words Frequency Timestamp Custom_Name_GenAI
0 0 vendor, weed, xanax, cocaine, mg 1563 2019-08-05 17:55:40.800 Drug Vendor Europe
1 0 vendor, weed, xanax, mg, cocaine 7054 2019-10-29 08:00:00.000 Drug Vendor Europe
2 0 vendor, nt, weed, drug, looking 28 2020-01-21 16:00:00.000 Drug Vendor Europe
3 1 market, empire, dream, vendor, nightmare 1102 2019-08-05 17:55:40.800 Dream Vendor Nightmare
4 1 market, empire, dream, vendor, nightmare 5109 2019-10-29 08:00:00.000 Dream Vendor Nightmare
In [17]:
sg.plot_avg_prob_or_freq(topics_over_time, 'Frequency')
No description has been provided for this image
In [20]:
sg.plot_boxplot(topics_over_time, 'Frequency')
No description has been provided for this image

Visualize Graphs¶

In [7]:
topic_model = BERTopic.load("../../../Analyze_files/CombiningAnalysisCompleteDataset/ThreadAnalysis/Models/topic_model_0.64SilNew", embedding_model='distiluse-base-multilingual-cased-v1')

image.png

7ClusterDistribution_0.64sil1200_after.png

7MatrixDistribution_0.64sil1200_LLAMA.png

7HierachicalTimeSeries_0.641200_LLAMA.png

In [8]:
topic_model.visualize_barchart(top_n_topics=10, custom_labels=True, n_words=10, width=350, height=350)

image.png

In [9]:
topic_model.visualize_topics_over_time(topics_over_time, top_n_topics=10, width=1250, height=700, custom_labels=True)

image.png

Predict Models - Usage¶

In [12]:
topic_model = BERTopic.load("../../../Analyze_files/CombiningAnalysisCompleteDataset/ThreadAnalysis/Models/topic_model_0.64SilNew", embedding_model='distiluse-base-multilingual-cased-v1')
In [4]:
sentence = ['i want to by some xanax and guns and coke']
topic, _ = topic_model.transform(sentence)
topic_model.get_topic_info(topic[0])
Out[4]:
Topic Count Name CustomName Representation Representative_Docs
0 0 8645 0_weed_xanax_vendor_cocaine Drug Vendor Europe [weed, xanax, vendor, cocaine, mg, uk, coke, b... [high quality weed thc product europe, new xan...
In [5]:
sentence = ['really bad vendors he scammed me']
topic, _ = topic_model.transform(sentence)
topic_model.get_topic_info(topic[0])
Out[5]:
Topic Count Name CustomName Representation Representative_Docs
0 2 6907 2_vendor_scammer_scam_looking Trusted Vendor Scams [vendor, scammer, scam, looking, scamming, sal... [looking good reliable vendor sell ounce, vend...
In [6]:
sentence = ['xtc']
topic, _ = topic_model.transform(sentence)
topic_model.get_topic_info(topic[0])
Out[6]:
Topic Count Name CustomName Representation Representative_Docs
0 4 6299 4_mdma_lsd_get_looking Drug Discussion [mdma, lsd, get, looking, wsm, good, btc, ques... [hey ro im gon na pull pk, life wonderful life...
In [7]:
sentence = ['closed market']
topic, _ = topic_model.transform(sentence)
topic_model.get_topic_info(topic[0])
Out[7]:
Topic Count Name CustomName Representation Representative_Docs
0 1 6236 1_market_empire_dream_nightmare Dream Vendor Nightmare [market, empire, dream, nightmare, vendor, dre... [next market, dream market vendor rstclass nig...
In [8]:
sentence = ['i want to buy coke cutted']
topic, _ = topic_model.transform(sentence)
topic_model.get_topic_info(topic[0])
Out[8]:
Topic Count Name CustomName Representation Representative_Docs
0 0 8645 0_weed_xanax_vendor_cocaine Drug Vendor Europe [weed, xanax, vendor, cocaine, mg, uk, coke, b... [high quality weed thc product europe, new xan...

HuggingFace¶

In [ ]:
topic_model.push_to_hf_hub(
    repo_id="D0men1c0/ISSR_Dark_Web_7Topics",
    save_embedding_model='distiluse-base-multilingual-cased-v1',
    save_ctfidf=True
)
In [16]:
topic_model = BERTopic.load("D0men1c0/ISSR_Dark_Web_7Topics")
In [17]:
sentence = ['closed market']
topic, _ = topic_model.transform(sentence)
topic_model.get_topic_info(topic[0])
Out[17]:
Topic Count Name CustomName Representation Representative_Docs
0 1 6236 1_market_empire_dream_nightmare Dream Vendor Nightmare [market, empire, dream, nightmare, vendor, dre... NaN